How to rank in AI search results in 2026
Last updated 28 October 2026

Guide
AI search doesn't rank pages, it picks passages it can retrieve, extract and verify. How to pass all three gates, with a 30-day plan for small teams.
To rank in AI search results, make sure AI crawlers can reach your pages, write passages that answer a question on their own, and get other credible sources to mention you in the same context. There is no ranking position to chase, only inclusion in the answer.
The key message: AI search does not rank pages, it selects passages it can retrieve, extract and verify. Every tactic that works falls under one of those three gates, and most startups fail at the first or the third without noticing.
What "ranking" means in AI search
Traditional SEO orders ten blue links. AI search tools (ChatGPT search, Perplexity, Gemini, Google AI Overviews) write a synthesized answer and attach a handful of sources. Being one of those sources is the new ranking.
Three terms are worth defining, because they get blurred:
- GEO (generative engine optimisation): shaping your site and your wider web presence so generative AI tools cite or recommend you.
- AEO (answer engine optimization): structuring content so it can be lifted directly as an answer. Our guide to AEO covers the differences from SEO and GEO.
- Citation: a linked or named source inside an AI-generated answer. A mention is the brand name appearing in the answer without a link. Both matter, and mentions often come first.
SEO still matters. Google AI Overviews draw on pages that Google has indexed, and many AI tools lean on search indexes. But a page that ranks fifth on Google can lose to a page ranked twelfth if the lower page contains a cleaner, more verifiable answer.
The three-gate model
When someone asks an AI tool "what's the best invoicing tool for freelancers," roughly this happens. The system turns the question into one or more searches, retrieves candidate pages, pulls out passages, and writes an answer from the ones it trusts. You can lose at any step.
| Gate | Question it answers | Typical failure |
|---|---|---|
| Access | Can the system fetch and read the page? | Blocked crawlers, JavaScript-only content, firewall rules |
| Extraction | Can it lift a clear, self-contained answer? | Vague intros, answers buried, claims without specifics |
| Corroboration | Do other sources agree with or repeat this? | No third-party mentions, inconsistent descriptions |
Most advice online focuses on gate two. Gate one is cheaper to fix and gate three is where competitors usually pull ahead.
Gate 1: make your site accessible to AI crawlers
If a crawler cannot fetch your page, nothing else matters. Check these in order.
Review robots.txt for AI user agents
Different bots do different jobs. OpenAI uses GPTBot (associated with training) and OAI-SearchBot (used for search results). Anthropic uses ClaudeBot. Google uses Googlebot for Search, and Google-Extended is a separate control for certain AI training uses. Blocking Google-Extended does not remove you from AI Overviews, which depend on normal Search eligibility.
Many sites block "AI bots" with a blanket rule copied from a template and then wonder why they never appear in answers. Decide deliberately: if you want citations from a given tool, its search crawler needs access. You can check specific pages with the free robots.txt tester.
Check the layers robots.txt does not cover
A permissive robots.txt does not help if your CDN, WAF or bot-protection rules return a 403 to unfamiliar user agents. Test by requesting a page with the crawler's user-agent string and looking at the response code, not just the file.
Serve real HTML
Many AI crawlers do not execute JavaScript well, or at all. If your pricing, feature list or core copy only appears after client-side rendering, view the page source. If the text is not there, server-render it.
Keep pages indexable
Google states that there are no special technical requirements for appearing in AI Overviews beyond being indexed and eligible to show a snippet. A stray noindex, a nosnippet directive, or a canonical pointing elsewhere removes you. Audit those before writing anything new.
Gate 2: write passages that can be extracted
AI systems cite passages, not pages. A passage is a few sentences under a heading that makes sense without the rest of the article. Write every important section as though it will be read in isolation.
Lead with the answer
Put the direct answer in the first one or two sentences under a heading, then add nuance. If your section "How much does X cost?" opens with a paragraph about how pricing is complicated, the extractable sentence is somewhere in paragraph three, and a competitor's page that opens with a number will be chosen instead.
Name things precisely
Compare two sentences:
- "Our tool helps teams work more efficiently."
- "Acme is a time-tracking app for agencies with fewer than 20 people that exports directly to QuickBooks."
The second can be quoted in answer to "time tracking app for small agencies." The first can be quoted in answer to nothing. Include the category, the audience, the key constraint and the differentiator in plain words, and keep that description identical across your homepage, docs, directory listings and social profiles.
Cover the sub-questions
AI tools often expand one question into several. "Best project management tool for startups" turns into questions about price, integrations, learning curve and team size. A page that answers those sub-questions under their own headings gives the system more reasons to retrieve it. Our guide on getting cited by Google AI Overviews walks through this process.
Use original, checkable detail
Generic summaries of what everyone already says are interchangeable, so they are rarely cited. Things models cannot get elsewhere include your own benchmark, a pricing comparison you compiled, a dataset from your product (anonymised), or a specific teardown. State the method and the date. If you cite a statistic, name the source.
Format for scanning
- Question-style H2 and H3 headings that match how people phrase prompts.
- Short paragraphs, with one idea each.
- Tables for comparisons.
- A short FAQ with direct answers, 40 to 80 words each.
- Visible publish and update dates. Stale pages lose to fresh ones on fast-moving topics.
Structured data (Article, FAQPage, Organization) is cheap to add and helps machines identify what the page is. Treat it as support for good content, not a substitute.
Gate 3: earn corroboration outside your site
A model asked to recommend a product is cautious about repeating claims that only the vendor makes. Corroboration is what turns "this company says it is the best" into "multiple sources describe it this way." This is the gate most founders skip, because it cannot be done by editing your own site.
Reddit threads regularly appear both in Google results and as sources in AI answers, particularly for "best X for Y" and "X vs Y" queries. That makes Reddit one of the few places where a founder can directly influence the evidence an AI tool reads.
What works:
- Find threads where your category is already being discussed. Search for the questions your buyers ask, and look at threads that rank on Google, since those are the ones most likely to be retrieved.
- Answer the actual question first. A reply that solves the problem and mentions your product as one option, with honest trade-offs, survives moderation and gets upvoted.
- Disclose that you built it. Undisclosed promotion gets removed and damages the account.
- Write titles like search queries when you start a thread. The free Reddit title analyzer scores a draft title and suggests rewrites.
- Return to threads. Old threads keep getting retrieved, so a useful reply compounds.
Geograph's Monitor and Reply tools exist for this workflow: Monitor surfaces threads that match your category, and Reply drafts answers you edit before posting. Whatever tool you use, a human should read every reply before it goes live.
Contextual backlinks
Links still signal trust to search engines, and AI tools that use search indexes inherit that signal. What matters is context: a link from an article in an adjacent niche that discusses the same problem carries more weight than one from an unrelated directory. Quantity is a poor target. Our tutorial on building backlinks for AI search covers the steps in detail.
Third-party mentions
Get listed accurately in places that compile comparisons: relevant directories, "best of" roundups, integration partner pages, podcasts and newsletters in your niche. Each is another independent source repeating the same description of what you do. If those descriptions contradict each other, models will either blend them badly or skip you.
How to measure whether it is working
AI visibility has no Search Console equivalent that covers every tool, so build a simple manual process.
- Write a fixed prompt set. Pick 15 to 25 prompts your buyers would plausibly use, including "best [category] for [audience]," "[you] vs [competitor]," and problem-led questions.
- Run them monthly in ChatGPT, Perplexity, Gemini and Google (for AI Overviews). Log whether you are cited, merely mentioned, or absent, and who appears instead.
- Note the cited sources. When a competitor wins, look at which pages were cited. That tells you which gate you are losing at.
- Watch referral traffic from AI tools in analytics, but expect it to be small and incomplete. Some answers produce brand searches rather than clicks.
- Track Google rankings and Reddit referrals alongside, since all three feed each other.
Expect results to vary between runs, because AI answers are not deterministic. Look at trends across many prompts, not single outputs. The SERP visibility guide lists further metrics worth tracking.
A 30-day plan for a small team
- Week 1: access. Audit robots.txt, firewall rules, rendering and indexation on your ten most important pages. Fix blockers.
- Week 2: extraction. Rewrite those ten pages: answer-first openings, precise product description, FAQ, visible dates. Add one original data point or comparison to each.
- Week 3: corroboration. Standardize your one-sentence description across every profile. Find ten Reddit threads that rank for your category and post useful, disclosed replies in the ones where you have something to add.
- Week 4: baseline and gaps. Run your prompt set, record results, and pick the three prompts where a competitor wins. Look at what sources were cited and plan content or outreach to match.
Common mistakes
- Blocking all AI bots by default and then expecting citations.
- Publishing volume instead of substance. Fifty thin posts give models nothing to quote.
- Burying the answer under an introduction.
- Ignoring Reddit because it feels off-brand, while competitors shape the threads AI tools read.
- Inconsistent positioning. Calling yourself three different things across the web.
- Buying links in bulk with no topical relevance.
- Checking once. AI outputs shift, so a single test tells you very little.
The takeaway
Pick the three prompts where a competitor is recommended and you are not, open the pages those answers cite, and ask which gate you fail: can the bot reach you, can it quote you, does anyone else back you up? Fix that one gate for those three prompts before writing anything new. If you want the monitoring, Reddit replies, article drafting and audits in one place, Geograph runs them as a connected system, so what you learn from the prompts that are not citing you feeds straight into what you publish next.
Frequently asked questions
- Is GEO different from SEO?
- They overlap heavily. SEO aims to rank pages in a results list. GEO aims to be included in a generated answer. Technical health, quality content and links help both, but GEO adds emphasis on extractable passages, third-party mentions and Reddit.
- Do I need to rank on Google to be cited by AI tools?
- Not always, but it helps a lot. Google AI Overviews draw from Google's index, and several other tools rely on search results to retrieve sources. Ranking raises your chances but does not guarantee a citation.
- How long does it take to appear in AI answers?
- Fixing access and rewriting key pages can show up within weeks for tools that retrieve live pages. Corroboration through Reddit and links builds more slowly, typically over a few months. There is no guaranteed timeline.
- Should I allow GPTBot and other AI crawlers?
- That is a business decision. If you want to be cited in a given tool, allow its search crawler (for OpenAI, OAI-SearchBot). Training-related crawlers are a separate choice, and you can treat them differently.
- Is Reddit worth it for a small startup?
- Yes, if you participate honestly. A few helpful, disclosed replies in threads that already rank can influence both Google results and AI answers, and the cost is time rather than money.
- Can I trust AI-generated content to rank?
- Only if a human adds something the model could not: original data, real experience, accurate specifics. Unedited generic output is interchangeable with every other site's, which is the main reason it does not get cited.
Start getting found on AI, Google and Reddit
- 15 articles a month optimised for AI & search citation
- Get backlinks on autopilot from other members
- Real time Reddit monitoring with reply drafts to drive traffic and AI mentions
- Weekly deep site audits with a full report card, tracked over time
- All built to maximise AI mentions and search ranking
Pricing and figures mentioned are accurate as of publish (28 October 2026) and may have changed since - check each provider's site for current numbers.
Written by Toby Marshman

